Glossary

Image Pixel Module

(0028,0002) Samples per Pixel

The number of samples per pixel, otherwise known as the number of image channels, components or planes. An RGB image has 3 samples per pixel (red, green and blue), a grayscale image has 1 sample per pixel (intensity). The Samples per Pixel for all DICOM Pixel Data is either 1 or 3, however 4 was previously allowed.

Allowed values: 1 or 3, but may be constrained by the IOD.

(0028,0004) Photometric Interpretation

The intended interpretation of the Pixel Data in its current form in the dataset. For example:

  • If you have a dataset with RGB Pixel Data then the Photometric Interpretation should be 'RGB'.

  • If you take your RGB data and convert it to YCbCr then the Photometric Interpretation should be 'YBR_FULL' (or a related interpretation depending on the conversion method).

  • If you then compress that data using RLE Lossless encoding then the Photometric Interpretation remains 'YBR_FULL'.

  • On the other hand, if you take your original RGB data and apply JPEG 2000 Lossless encoding then the Photometric Interpretation will either be 'RGB' or 'YBR_RCT' depending on whether or not the encoder performs multiple component transformation.

When compressing pixel data using one of the JPEG encodings it’s important to know if the encoder is performing any color space transformation prior to compression, as this needs to be taken into account when setting the Photometric Interpretation. This is important when an encoder performs a transformation and the decoder doesn’t, since having a correct Photometric Interpretation makes it possible to determine which inverse transformation to use to return the pixel data to its original color space.

For more detailed information on each of the defined photometric interpretations refer to Annex C.7.6.3.1 of Part 3 of the DICOM Standard.

Allowed values: 'MONOCHROME1', 'MONOCHROME2', 'PALETTE COLOR', 'RGB', 'YBR_FULL', 'YBR_FULL_422', 'YBR_PARTIAL_420', 'YBR_ICT', 'YBR_RCT', however restrictions apply based on the Transfer Syntax UID, and further constraints may be required by the IOD.

(0028,0006) Planar Configuration

Required when Samples per Pixel is greater than one, this indicates the order of the samples used by the pixel data, as either:

  • 0, where sample values for the first pixel is followed by the sample value for the second pixel: R1, G1, B1, R2, G2, B2, …, Rn, Gn, Bn.

  • 1, where sample values for each color plane are contiguous: R1, R2, …, Rn, G1, G2, …, Gn, B1, B2, …, Bn.

Allowed values: 0 or 1

(0028,0100) Bits Allocated

The number of bits used to actually contain each sample of each pixel. All DICOM Pixel Data is either 1 (for bit-packed Pixel Data) or more typically a multiple of 8 such as 8, 16 or 32, with 64 currently being the maximum used. Using the example of a Bits Stored of 12, this means that the actual number of bits used to contain the values must be at least 16.

For more detailed information refer to Chapter 8 and Annex D in Part 5 of the DICOM Standard.

Allowed values: 1 or a multiple of 8, however many IODs place further restrictions on what the value may be.

(0028,0101) Bits Stored

The number of bits actually used by each sample of each pixel. For example, with a Bits Stored value of 12, an unsigned grayscale image will have pixel values in the range 0 to 4095 and an unsigned RGB image will have values in the range (R: 0 to 4095, G: 0 to 4095, B: 0 to 4095). Must be equal to or less than Bits Allocated.

For more detailed information refer to Chapter 8 and Annex D in Part 5 of the DICOM Standard.

Allowed values: 1 to Bits Allocated (inclusive)

(0028,0102) High Bit

The most significant bit of the pixel sample data and is equal to Bits Stored - 1, however other values have been allowed in past versions of the DICOM Standard.

Allowed values: Bits Stored - 1

(0028,0103) Pixel Representation

Describes the type of pixel values, either signed (using 2’s complement) or unsigned integers. A value of 0 indicates the Pixel Data contains unsigned integers while a value of 1 indicates it contains signed integers.

Allowed values: 0 or 1, but may be constrained by the IOD.